COMP.SGN.220 Project work

The project aims to classify different musical instruments using CNNs.

General Information

• IRMAS dataset was used to train, validate and test the system.
• development was done in Python with PyTorch framework

Usage

• extract IRMAS-TrainingData and IRMAS-TestingData-Part1 to the current workspace
• run serialize_data.py to preprocess the dataset
• run training.py which contains the training, validation and testing procedures

Project Status

Project is: complete

Room for Improvement

Implement multi classifier

Acknowledgements

This project was based on this paper (Detecting and Classifying Musical Instruments with Convolutional Neural Networks) from D. Hing and C. Settle. First goal was to convert the work from TensorFlow to PyTorch and then further optimize the project.